In server virtualization, the key is thinking of an application as a workload. A server with a hypervisor can run multiple, independent workloads. This allows for better utilization of the server and much greater flexibility, agility, and responsiveness in managing workloads.
In routing, we have seen many attempts to run multiple control plane workloads trying for years on legacy routers. VRF, VRF-lite, and node sliding are all options for legacy routers. As we have noted many times, legacy routers are appliances; they are tightly designed software and hardware bundles that meet a set of requirements. The CPU on a routing control module is sized to these requirements. Coupled with the sheer size of the codebases, it is difficult to see how multiple copies of the control plane software can run on a given router.
White box switches are gaining a great deal of interest. Projects like TIP’s DCSG actually specify the requirement for a white box switch. The reason is simple. The number of locations that require routing will increase by one to two orders of magnitudes. That means we will see the routing service edge grow from hundreds of PoPs to tens of thousands of cell sites over time. White boxes are the best way to drive costs down by the same degree.
These routers will require carrier-grade control plane software, so CPU demands are substantial if service providers continue with the appliance model by using Network Operating Systems (NOS). RAN sharing, serving enterprise customers and network slicing all benefit greatly from multiple virtual routers. However, the NOS model on low-cost hardware does have enough CPU to support multiple routing workloads. This is where the server analogy breaks down.
The best way to support multiple virtual routers is to embrace disaggregation. A limitation of the low-cost white box is the CPU. The cloud offers the best way to scale processing in the most cost-effective way possible. By moving the control plane processing to the cloud, Volta can support up to 255 virtual routers on a single switch. Thus, we can gain the benefits of lower costs and greater agility of multiple workloads by recognizing that the public, private or hybrid cloud represents the best way to scale.